Bug: MultiMatte RGB colors lower than 1

[sharktacos]

The RGB colors for a multimatte should have a value of 1 respectively. So image for the red matte should be <1,0,0> however the values seem to be typically less than 1, for example on one multimatte I get <0.9875,0,0> for the red, <0,0.956,0> for the green, and <0,0,0.819> for blue. Sometimes (I believe due to refractive objects) you can even get values NaN values like <1.919, 0, -1.3-e06> for a red channel matte.

This was tested in Maya 2015 with Vray 3.52, batch render, viewing the pixels in Nuke. I get similar results viewing the pixels in the VFB, although fwiw I don't see NaN numbers in the VFB.

A work around for this in Nuke is to isolate the channels in a shuffle, boosting its gain above 1, and then clamping it. However, this is obviously not the desired behavior from the render element.

Is this a known bug on the Chaos bug-tracker?

[vlado]

It seems to work fine here. Can you get me a scene that exhibits the issue?

Best regards,
Vlado

[sharktacos]

I've attached a zip with a Maya 2015 scene and two EXR renders. One with environment volume on, and another with it off. As you can see looking at the MM channel of both renders, the error appears to be connected to having the environment volume on.

[sharktacos]

Did you have a chance to look at the file? Is multimate breaking with environment volume a known bug?

[tashko.zashev]

Hi,

Thanks for the investigation you have done on this render problem.
I just want to clarify that having any semi transparent objects in front of the isolated geometry will affect it's color in the RGB and Multimatte pass. I don't think that we have an option or settings to sort that out, so the best workflow at the moment would be exactly what you described in the first post.

[francomanko]

Just for the record, i was wondering why my rgb values in the multimatte were not fully at 1. It was the vignette on the camera Would be good to have this not affect the maps maybe?